HSEM read lock vs write lock?

[AMurr.2282]

Hi!

So I was reading the wiki here https://wiki.st.com/stm32mpu/wiki/How_to_configure_system_resources

and they tell you to use PERIPH_LOCK, which is a function that can be found here:

https://github.com/STMicroelectronics/STM32CubeMP1/blob/1.3.0/Projects/STM32MP157C-DK2/Templates/Src/lock_resource.c

However, HAL_HSEM_FastTake does a read lock and not a write lock.

When should I be using HAL_HSEM_FastTake() vs HAL_HSEM_Take()? Why does PERIPH_LOCK call HAL_HSEM_FastTake() to get a read lock when it is writing to GPIO/EXTI registers?

[Olivier GALLIEN]

Hi @AMurr.2282​ ,

Reading the RM again, I understand you probably make a confusion between the 2 lock methods "label" (read) or (write) and the operation to protect.

Once semaphore is taken it lock peripheral registers whatever access type.

But there's 2 ways to lock :

– 2-step lock: by writing COREID and PROCID to the semaphore, followed by a read check -> labeled (write)

– 1-step lock: by reading the COREID from the semaphore -> labeled (read) method

I supposed advantage of "read" method is to be atomic.

Is that sound clearer ?

Olivier

[AMurr.2282]

> Is that sound clearer ?

Not really. In what situation would I need to use a write lock and why?

[Olivier GALLIEN]

Hi @AMurr.2282​ ,

Difference between 2-step (write) and 1-setp (read) is mentioned in RM :

"There is no PROCID available with the 1- step (read) lock" -> you can't identify the process which take the semaphore, only the COREID.

So, to arbitrate multi-process concurrency on same core, it can be more convenient to use the 2-step method.

Olivier